Guest Post by Willis Eschenbach
Anthony put up a post titled “Why the new Otto et al climate sensitivity paper is important – it’s a sea change for some IPCC authors” The paper in question is “Energy budget constraints on climate response” (free registration required), supplementary online information (SOI) here, by Otto et alia, sixteen other alia to be precise. I agree that it’s an interesting and important paper, mostly because of who the authors are. However, I have to say there were some parts that I couldn’t fit together. Here’s their figure S2, showing the radiative forcing since 1850:
Figure 1. Forcings used in Otto 2013. Grey lines represent possible variations based on a monte carlo analysis of the errors in the parts that make up the total radiative forcing. They divide the total forcings into greenhouse gases (GHGs), volcanoes and solar, and a “residual” forcing which is assumed to be mostly aerosols. The estimated errors in these are used to generate a host of possible realizations, shown in gray above.
Now, when I saw that, I just rolled my eyes. Here we go again with the volcanoes, I thought.
See those big dips in the black line above? Those are reckoned to be the change in forcing due to volcanic eruptions. What happens is that the volcanoes spew light-colored aerosols into the stratosphere. This reflects more sunlight, and thus reduces the forcing in a measurable manner. As you can see, the larger volcanoes make a very significant change in the forcing. So I set out once again to see if the claimed temperature change due to the volcanic forcing held up in the real world.
But as often happens, before getting to the volcanoes I got sidetractored, this time by discovering that Otto et al. to the sixteenth power are not discussing eruptions from ordinary active volcanoes . Oh, no indeed.
Otto and his hexadecagonic cohort are discussing retroactive volcanoes.
Here’s why I say that. I digitized the Otto data. Figure 2 shows the results, including the dates of the larger eruptions which are responsible for the large dips in the amount of sunlight reaching the earth:
Figure 2. Forcings from the Otto 2013 paper, along with the volcanoes associated with each of the large dips in forcing. The earliest eruption is assumed to be Cotopaxi because of the very large dust veil index associated with that eruption.
Here’s the oddity I noticed. As you can see, not only are the volcanoes associated with the large drops in forcing. They also apparently are able to cause the temperatures to drop during the year before they occur … in other words, they cool retroactively. In each case, there’s a drop in forcing, not only in the year of the eruption, but in the previous year as well …
Here’s the likely reason why the retroactive drop in forcing occurs. It’s not a timing error in the data. Figure 3 shows the Otto forcing data compared to the GISS forcing data.
Figure 3. Forcings from the Otto paper compared with the GISS forcings.
Note that in both the GISS and the Otto forcings, the lowest points after the eruptions coincide perfectly, so it’s not from timing … but the years immediately before the eruptions are quite different. Rather than the forcing falling in the year prior to the eruption as the Otto data does, the GISS forcing remains high until the actual year of the eruption. Additionally, the size of the GISS volcanic forcings is about twice the size of the Otto volcanic forcings.
Now, the paper says that they are using the average model forcings from the Forster paper that I discussed recently. However, the Forster forcings look much like the GISS forcing shown above, with large volcanic excursions. In addition, the Forster results are not retroactive—as with the GISS data, the Forster forcing drops in the year of the volcano, and not the previous year as in the Otto forcings.
I suspect (and let me emphasize suspect) that what has happened is that either rashly or quite possibly inadvertently, someone has slightly smoothed the Forster forcing data. The difference is subtle, I didn’t notice myself until I looked at Figure 2 and went whaa?
I say it’s smoothed because if I use a simple 3-year centered moving average on the Forster data, I get something very near to the Otto forcing data. Such a smoothing makes the volcanic drops “retroactive”, since the centered moving average includes the following year’s data. Figure 4 shows those results.
Figure 4. The Forster (blue) and Otto (red) forcing data, along with a smoothed version of the Forster data. The smoothing is done using a simple centered 3-year moving average. The years prior to the eruptions of Krakatoa (orange diamond) and Pinatubo (red squares) are highlighted in both datasets to show the “retroactive” effect of the smoothing on the previous year’s data. The Forster data has a 0.3 W/m2 trend added to match the Otto data, as described in the Otto paper SOI.
Note how similar the simple smooth of the Forster data is to the Otto data. This is the only explanation I can think of for the retroactive nature of the Otto volcanoes.
If the Otto data is indeed incorrect, that could make a large difference in their results. It’s difficult to say, but since the size of the volcanic excursions in forcing have been cut in half in the Otto data, it seems like it might be important. In addition, when you are using a forcing time series dataset for predicting (hind casting or forecasting) results, it’s a lot easier to forecast next years temperature data when your forcing data for this year contains some information about next year … using a smoothed dataset as input to another calculation is almost always a huge mistake.
For me, the best thing about the Otto paper was that via the Forster paper, it provided the data and the impetus to write my last post on climate sensitivity. It also provided an insight into how to analyze the effects of the volcanoes on the historical data versus the claims of the models regarding the volcanic effects … stay tuned, my new findings in that regard are the subject of my next post, interesting stuff.
Best wishes to all,
w.
DATA
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
How is the total radiative forceing found.
1. Is this solar plus gh radiation (in which case this graph clearly shows an anthopogenic signal).
2. Solar only (in which case it shows a solar signal).
3. Something else???
help appreciated
Willis
I am pleased to see that you have posted this article, it is an interesting read.
You open up stating: “Here’s their figure S2, showing the radiative forcing since 1850: Now, when I saw that, I just rolled my eyes. Here we go again with the volcanoes, I thought.” This was my reaction, and it appeared to me that something was not quite right. Unfortunately, I was rather lazy and I respect your analytical skills so I commented on your Model Climate Sensitivity Calculated Directly From Model Results article (richard verney says: May 22, 2013 at 2:39 am):–
“Willis
It would be useful to list the major volcanic eruptions since 1850 and the claimed negative feedback with respect to each.
Does anyone really think that Pinatubo (1991) had the same impact as Krakatoa (1883). Without digitalizing, to my unaided eye, the negative forcings appear similar.
Does anyone really have any confidence in the temperatures? Surely few people hold the view that today is about 0.6degC warmer than it was in the 1930s, or for that matter some 0.8degC warmer than the 1880s. I would have tjought that within the margins error, it is difficult to conclude that temperatures today today truly are warmer than those either in the 1930s or 1880s such that the present temperature record does not show the full extent of variability in past temperatures…..”
Thanks for dealing with my enquiry. Strange goings on indeed. I do consider that it is time for a detailed analysis of the effects of volcanoes on the temperature record/model outputs is required since I consider this an area where there is the potential for much fudging to try and assist back tuning.
Effect of land volcanoes on the atmosphere is temporary and minor except for the strongest eruptions.
Effect of submarine volcanic ‘slow eruptions’ in three critical areas (Arctic, Aleutian archipelago and Solomon Islands) in form of an ocean currents disturbance, may be far longer lasting and far more critical to the climate change.
I find it astonishing that anyone would use a time-symmetric filter for smoothing data to be used like this, precisely because of this effect.
If you HAVE to smooth data then using a backwards-only Savitsky-Golay filter would do the trick without introducing any forward-looking information content (see, for example, Numerical Recipes in C++, 3rd Ed. for details on constructing SG filters).
There is little excuse for such laziness of analysis, other, perhaps, than ignorance, which isn’t much of as excuse.
This seems pretty obviously to be a filter as Willis suggests. There is absolutely no reason to smooth volcanic forcing data so I don’t know why anyone would.
The question then is does it “matter”.
The area under the graph (the integral) should be the same in both cases, so it’s not going to change the energy in.
Does it make that resulting temperatures “look” more realistic. Possibly.
http://climategrog.wordpress.com/?attachment_id=258
If we look at the rate of change of CO2 at Mauna Loa as here (or other data such as SST , AO index ) we see that other climate variables drop BEFORE eruptions. This is purely coincidental but leads to a false attribution.
Now if modellers start spreading volcanic forcings back in time by smoothing them the results will correlate better with climate data that show variations prior to eruption that are nothing to do with volcanism.
This will lead to a spurious impression that the mode is somehow “capturing” real climate variation and correctly modelling the effect of volcanoes.
Since volcanoes are the pillar on which rests the exaggeration of CO2 forcing, making it look like the models fit the data better than they would if they did not probably IS significant.
Now we should always assume good faith so we should probably conclude that someone chose to smooth the volcanic data for no good reason , thus degrading the accuracy of the model but fortuitously this made the model output correlate better with the real climate data which was dropping before the eruption for other reasons.
The result of this unfortunate error is to make it look like the volcanic forcing is more accurate than it is which in turn supports pumping up CO2 with parametrised feedbacks.
Since that was the “expected” result anyway , this will not be a visible problem to the authors.
Greg Goodman says:
May 23, 2013 at 1:22 am
“…Now if modellers start spreading volcanic forcings back in time by smoothing them the results will correlate better with climate data that show variations prior to eruption that are nothing to do with volcanism.
This will lead to a spurious impression that the mode is somehow “capturing” real climate variation and correctly modelling the effect of volcanoes.
Since volcanoes are the pillar on which rests the exaggeration of CO2 forcing, making it look like the models fit the data better than they would if they did not probably IS significant…”
//////////////////////////////////////////////////////////////
This is exactly my concern as noted in my post of 10:20pm.
I am particularly concerned since do we really have reliable quantative data for any volcano eruption, let alone those before say the late 1970s?
How were we measuring what quantity of gases, aerosls and the like were emitted, at what height did these emissions take place, what were the jet streams at the relevant time etc.
The potential for fudging is simply enormous since the required data simply does not exist. All we can do is talk in generalties such as Krakatoa was a larger eruption thann Novarupta which in turn was larger than Pinatubo.
Further, how many volcanoes of note are simply excluded from the record (cannot be seen in the record), perhaps because if they were to be included then discrepancies would become apparent.
I really do consider that a summary of every major volcano eruption since 1850 should be listed with the ‘guestimates’ of the contents of the eruptions etc. I consider that a further review of the applied forcings in the light of that summary should be conducted.
“The potential for fudging is simply enormous since the required data simply does not exist.”
Indeed, all guestimates of how much of what got thrown how far is real back of envelop stuff.
However, what really needs closer examination is where there is any real evidence of a permanent negative shift in temperature after one of these eruptions. Every time I look at this, all I find is a not so certainly attributable negative glitch and NO permanent offset.
There has to be some radiation forcing during the time the ejections are in the atmosphere but I don’t see any evidence of a lasting effect. This implies that climate feedbacks correct for the perturbation.
Willis has already suggested a mechanism exists in tropical storms.
Would not the eruption latitude also make a difference as would hemisphere. There is little mixing between hemispheres and the more polar the eruption the less insolation would be reduced by the SO2.
Willis has already suggested a mechanism exists in tropical storms. …
Tropical storms as emergent phenomena are a negative feedback with it’s own internal positive feedback (making it a more and more powerful negative f/b). These are fast responding (hourly) phenomena that reduce local hotspots.
These storms are too small-scale to be reproduced by climate models. So instead of the basic physics built into the model producing the feedback, parameters are guessed and plugged into the models as an inputs.
Now although TS are local and fast they can also provide a correction for a perturbation like a tropical eruption.
If waters are cooler there will be less local hotspots that set the local conditions that trigger a storm (and vice versa). This will be a reaction to generally cooler surface conditions and thus will react to annual or inter-annual changes in SST, not just hourly local conditions.
Now such a mechanism will not simply act as a simple negative feedback that will compensate for the lack of insolation , restoring equilibrium energy balance, and leaving a new lower equilibrium temperature. That is what the model behaviours suggest happens.
What TS feedback will do is ensure a return to the _previous_ equilibrium SST. This is a stronger response than a simple linear negative feedback. They can do this because of the internal positive feedback that makes them into NON-linear negative feedbacks.
I think this is what Willis has been trying to get over with his “governor” concept.
I have had trouble getting his point because of his insistence that they are “not feedbacks” and the imprecision of the term “governor”.
I think this description of TS feedback , which could be modelled regionally on a smaller scale than current GCMs, would better fit the real climate data than the permanent temperature offset produced by current modelling efforts.
It is not so much the volcanic forcing that is wrong but the _assumed_ climate response.
Most likely this strange phenomenon is caused by bad statistical practices, something not unknown in climate science.
But let’s suppose for a moment that climate responses can indeed occur in the year preceding a major eruption.
One possible explanation did occur to me. Is it possible that there is a global increase in volcanic activity during the year running up to a major eruption? If so, then the climate responses could be caused by a general, but small, increase in global volcanic activity. It would be the equivalent of playing the overture before the main opera.
Chris
Did the eruption of the volcanoes in Iceland in 2010 not have any impact at all? There doesn’t seem to be anything showing on any of the graphs.
No, I think it is clear enough from that plot that this is the same data filtered with a 5y running mean. (Running mean because of the way it fails to follow in 1905 and 1960; 5y because of the width of the feature where it fails.)
Philip Aggrey says:
Did the eruption of the volcanoes in Iceland in 2010 not have any impact at all? There doesn’t seem to be anything showing on any of the graphs.
No. too small and too high latitude.
@ur momisugly Greg Goodman
The degree of coherence between volcanic & climate indices suggests coupling. A mathematical analogy is the balance of wheel rotation acceleration across a differential with turning, but here there are multiple wheels running off each axle and the turning is in more than 2 dimensions. The universal constraints are what matter globally and they can be derived. Combining points made by Bill Illis (May 22, 2013 at 9:03 pm) with points you’ve made about false mainstream assumptions, it looks to me like we potentially have roots of convergent thinking …but there’s a big However: at some timescales the coupling most certainly does not have the appearance of unidirectionality — that’s where I see much of the commentary at WUWT about volcano-climate relations being at odds with multivariate observational records. The data clearly suggest an external governor of both volcanism & climate indices. This isn’t a point I would debate. Rather it’s a point I assert.
Willis, I’ve just seen the caption on your figure 4
The smoothing is done using a simple centered 3-year moving average
Try 5 year and I’ll bet your bang on.
“your” illiterate Goodman !
I tried to find good volcano data on line, and didn’t, as someone already posted, lots of big volcano’s (on eruption scale) don’t seem to show up in the temp data, where smaller scale ones do. Though I do think some of it is where the volcano is. I think part of it is Global effect vs a more regional effect.
Paul Vaughan says: The degree of coherence between volcanic & climate indices suggests coupling….The data clearly suggest an external governor of both volcanism & climate indices. This isn’t a point I would debate. Rather it’s a point I assert.
Could you be more specific about where this ‘coherence’ can be found?
My comments above are based on looking at annual to inter-annual scale. That is not necessarily at odds with what you are suggesting which would presumably be on inter-decadal scales.
If you are aware of any short term coherence I would also be very interested as to where this can be seen.
Warmist strategy:
1. Play the arctic-ice-card
2. Play the volcano-card
3. Play the aerosol-card
4. Play the what-about-the-grandchildren-card
5. Play the mother-earth-card
Pamela Gray says:
May 22, 2013 at 6:04 pm
“Hey…animals can detect an earthquake before humans can. ”
This, and the often reported high pitched sound ahead of an earth quake. The shock wave (sound wave in rock) travels some 5000ft/second and would be felt and heard. The only thing I can think of that could get to one’s ears quicker would be electromagnetic. I’ve wondered if piezoelectric discharges from quartz in the rocks being strained, resulted in radio waves and surrounding structures acting as a receiver. I’ll probably be up for abuse again at raising this!
Paul Vaughan: The data clearly suggest an external governor of both volcanism & climate indices. This isn’t a point I would debate. Rather it’s a point I assert.
I’m quite interested in that line of investigation but I’d be a lot more interested in evidence than assertion. You seem pretty confident about it so I’m sure you’ve got something interesting.
[BTW , I would like to ask you about a comment you made on persistence of 9y cycles and your wavelet analysis, that you made recently if you would contact me http://climategrog.wordpress.com ]
“Here’s the oddity I noticed. As you can see, not only are the volcanoes associated with the large drops in forcing. They also apparently are able to cause the temperatures to drop during the year before they occur … in other words, they cool retroactively.”
It is quite usual that land temperatures are lower in the run up to the events than afterwards:
http://wattsupwiththat.com/2012/07/30/new-data-old-claims-about-volcanoes/#comment-1050905
Also, what is the evidence for a large eruption at Cotopaxi in 1855?:
http://www.volcanolive.com/cotopaxi.html
Well that is that sewn up. What are the reviewers doing? Shows you how flawed the peer reviewed process is. Obvious but takes a sharp eye to see it.
I suspect this is rather at the level of a knit pick than a grave error that should have been spotted by review process.
Perhaps Willis could plot the difference between the unfiltered data and the 5y smoothed version so we can see directly how big this is. The spread before Mt Agung in 1963 was because of other activity apparently not the filtering.
Deconvolving components of times series data is model dependent, so one should try different models to get an estimate of the model systematic errors [biases] and included this in the overall uncertainty.